Re: [HTCondor-users] matching gpu devicename

["John (TJ) Knoeller" <johnkn@xxxxxxxxxxx>]

These are effectively the same device?  It's annoying that they have 
separate names then. You should complain to NVidia... ;)

But seriously, your best course for now would be to condor_gpu_discovery 
program in a script that re-writes it's output
to change the various CUDA[n]DeviceName attributes to a single 
CUDADeviceName = "Tesla K10" attribute.

Or tell your users to target the GPUs based on capability rather than by 
name.

If the '.' in the name a standard pattern for GPU names and if we can 
reasonably assume that
everything after the first '.' in the name is noise, then 
condor_gpu_discovery  tool
could be changed to truncate the names to everything before the first 
'.' before publishing the name.

-tj

On 6/1/2015 8:24 AM, Michael Di Domenico wrote:
> How does one match a slot on a machine where the gpu's are mixed by devicename?
>
> for example
>
> if i have a machine with 6 slots where
>
> slot1-4 are
> CUDA0DeviceName = Tesla K10.G2.8GB
> CUDA1DeviceName = Tesla K10.G2.8GB
> CUDA2DeviceName = Tesla K10.G2.8GB
> CUDA3DeviceName = Tesla K10.G2.8GB
> slot5-6 are
> CUDA4DeviceName = Tesla K10.G1.8GB
> CUDA5DeviceName = Tesla K10.G1.8GB
>
> because i have multiple GPU device name types in the machines there is
> no global CUDADeviceName classad.
>
> if i do
>
> condor_status -constraint 'regexp("Tesla K10.G2.8GB",
> CUDA0DeviceName)' everything works out fine
>
> but is there a way to find all the cards in the pool regardless of
> which CUDA<num> they are?
>
> this will specifically apply to us because we have users that write
> cuda code optimized for a specific gpu and they'll want to put in
> their requirements expression regexp("Tesla K10", CUDADeviceName)
> which doesn't match anything currently
>
> the difference between the two cards above is airflow direction , not
> technical, so they're really the same card
> _______________________________________________
> HTCondor-users mailing list
> To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users
>
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/htcondor-users/